GRASP: GRanularity-Aware Search Policy for Agentic RAG

Hugging Face Daily Papers Papers
agentic-rag reinforcement-learning multi-hop-reasoning retrieval granularity-aware search-policy

Summary

Introduces GRASP, a reinforcement learning framework that trains agents to adaptively coordinate semantic search, keyword search, and paragraph reading during multi-step reasoning, improving retrieval recall and question answering performance on multi-hop benchmarks.

Agentic retrieval-augmented generation (RAG) extends static RAG by allowing language models to iteratively reason, generate search queries, retrieve evidence, and predict answers. However, it remains challenging for models to decide when to retrieve, whether to use lexical matching or semantic similarity, and how to control context granularity to prevent irrelevant tokens from interfering with agent reasoning. In this paper, we introduce GRASP, a reinforcement learning (RL) framework for training agents to adaptively coordinate complementary retrieval tools during multi-step reasoning. GRASP provides the agent with semantic search, keyword search, and paragraph-reading actions, enabling it to retrieve sentence-level evidence and expand further context only when needed. We train the policy with a reward that jointly accounts for answer accuracy, grounded reading, complementary search, and turn efficiency. Experiments on multi-hop reasoning benchmarks show that GRASP improves both retrieval recall and downstream question answering performance compared with single-step retrieval, prompting-based agentic RAG, and RL-based retrieval baselines. Qualitative and ablation analyses show that the learned policy develops interpretable skimming and scanning behavior: it uses semantic search for broad exploration, paragraph reading for local verification, and keyword search for entity-specific evidence. These results suggest that learning to coordinate retrieval signals and context granularity is critical for agent's correct reasoning.
Original Article
View Cached Full Text

Cached at: 07/20/26, 09:42 AM

Paper page - GRASP: GRanularity-Aware Search Policy for Agentic RAG

Source: https://huggingface.co/papers/2607.10463

Abstract

Agenticretrieval-augmentedgeneration(RAG)extendsstaticRAGbyallowinglanguagemodelstoiterativelyreason,generatesearchqueries,retrieveevidence,andpredictanswers.However,itremainschallengingformodelstodecidewhentoretrieve,whethertouselexicalmatchingorsemanticsimilarity,andhowtocontrolcontextgranularitytopreventirrelevanttokensfrominterferingwithagentreasoning.Inthispaper,weintroduceGRASP,areinforcementlearning(RL)frameworkfortrainingagentstoadaptivelycoordinatecomplementaryretrievaltoolsduringmulti-stepreasoning.GRASPprovidestheagentwithsemanticsearch,keywordsearch,andparagraph-readingactions,enablingittoretrievesentence-levelevidenceandexpandfurthercontextonlywhenneeded.Wetrainthepolicywitharewardthatjointlyaccountsforansweraccuracy,groundedreading,complementarysearch,andturnefficiency.Experimentsonmulti-hopreasoningbenchmarksshowthatGRASPimprovesbothretrievalrecallanddownstreamquestionansweringperformancecomparedwithsingle-stepretrieval,prompting-basedagenticRAG,andRL-basedretrievalbaselines.Qualitativeandablationanalysesshowthatthelearnedpolicydevelopsinterpretableskimmingandscanningbehavior:itusessemanticsearchforbroadexploration,paragraphreadingforlocalverification,andkeywordsearchforentity-specificevidence.Theseresultssuggestthatlearningtocoordinateretrievalsignalsandcontextgranularityiscriticalforagent’scorrectreasoning.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2607\.10463

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2607.10463 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.10463 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.10463 in a Space README.md to link it from this page.

Collections including this paper1

Similar Articles

GRACE-RAG: Governed Retrieval Architecture for Canonical Evidence Synthesis, Enabling Lightweight Deployment in Closed-Domain Institutional Settings

arXiv cs.AI

This paper introduces GRACE-RAG, a retrieval-governed, graph-augmented RAG architecture that externalizes structural reasoning from generation to a structured retrieval layer, enabling lightweight deployment in closed-domain institutional settings. Experiments show up to 20% quality gains with mid-scale models, reducing computational and latency footprint.